Abstract Robot-based thermal spraying is a production process in which an industrial robot guides a spray gun along a path in order to spray molten material onto a workpiece surface to form a coating of desired thickness. This paper is concerned with optimizing a given path of this sort by post-processing. Reasons for doing so are to reduce the thickness error caused by a not sufficiently precise design of the given path, to adapt the path to a changed spray gun or spray technology, to adapt the path to slight incremental changes of the workpiece geometry, or to smooth the path in order to improve its execution by the robot. An approach to post-optimization using the nonlinear conjugate gradient method is presented which employs a high-quality GPGPU-based simulation of the spray process for the evaluation of the coating thickness error and additionally taking care of the kinematic path quality. The number of computationally time-consuming calls of the simulation is kept low by analytically calculating estimates of gradients from a simplified material deposition model. A rigorous experimental evaluation on case studies of the mentioned applications shows that the method efficiently delivers improved paths which reduce the coating error on real free form surfaces considerably, i.e. the squared coating error is below 3.5% of the original value in every case study.
Read full abstract